How to Connect Claude to Live Options Data (MCP Server Setup Guide)

Large language models are good at reasoning about markets and bad at knowing what the market is doing right now. The Model Context Protocol (MCP) closes that gap: it lets an AI assistant like Claude call real tools against live data instead of guessing. This guide connects Claude to institutional-grade options analytics — IV rank, dealer gamma exposure, expected moves, screeners, strategy builders — in about two minutes.

What is an MCP server?

MCP (Model Context Protocol) is an open standard for wiring AI assistants to external tools and data. An MCP server advertises a set of typed tools; the assistant decides when to call them and interprets the results. The ApexVol MCP server (apexvol-mcp on PyPI) runs locally on your machine and talks to apexvol.com over HTTPS with your API token — it exposes 43 options-analytics tools across 8 categories, backed by 59 documented REST endpoints.

The practical difference: ask a bare model “Is NVDA’s earnings straddle overpriced?” and you get plausible-sounding text. Ask a model connected to an options MCP server and it calls an earnings-move tool, compares the implied move against the realized history, checks IV rank, and answers from numbers it just fetched.

What you can ask once connected

• “What’s the IV rank on SPY, and is the vol risk premium positive?”
• “Where’s the gamma wall on QQQ this week?”
• “Should I buy the AMZN earnings straddle or sell it?”
• “Screen for names with IV rank above 80 and no earnings in the next two weeks.”
• “Build an iron condor on SPY around the expected move and stress it ±10%.”
• “Pull the 30-delta put on NVDA and calculate its probability of profit.”

The 8 tool categories: options chain, volatility, Greeks & GEX, options flow, strategy, risk, events & screening, and ticker analytics. The complete tool list lives on the API & MCP page.

Before you start

• Python 3.10 or newer on your machine.
• An ApexVol Pro subscription.
• An API token from Account → API Access — tokens look like avmcp_…, are shown once at creation, and can be rotated in one click.

Step 1 — Install from PyPI

pipx keeps the command isolated and on your PATH (plain pip install apexvol-mcp also works):

pipx install apexvol-mcp

Step 2 — Verify the install and your token

Before touching any Claude config, confirm the command works and the token authenticates. This prints the client version, the auth result, and your remaining quota:

APEXVOL_API_TOKEN=avmcp_YOUR_TOKEN apexvol-mcp --check

Step 3 — Connect Claude

Claude Code (CLI) is one command:

claude mcp add apexvol -e APEXVOL_API_TOKEN=avmcp_YOUR_TOKEN -- apexvol-mcp

Claude Desktop: add this to claude_desktop_config.json (macOS: ~/Library/Application Support/Claude/; Windows: %APPDATA%\Claude\), then restart the app:

{
  "mcpServers": {
    "apexvol": {
      "command": "apexvol-mcp",
      "env": { "APEXVOL_API_TOKEN": "avmcp_YOUR_TOKEN" }
    }
  }
}

Pasting your token into a local config file is the intended way to connect. If a token is ever exposed somewhere public, rotate it under Account → API Access — the old secret dies instantly and the label carries over.

Prefer raw REST?

Everything the MCP tools can do, curl can do — same token, same endpoints:

curl -H "Authorization: Bearer avmcp_YOUR_TOKEN" \
     "https://apexvol.com/api/mcp/data/iv-rank/NVDA"

See the options data API guide for endpoint families, conventions and limits, or the OpenAPI 3.1 spec in the API documentation.

Access and limits

API & MCP access is included with the Pro plan. Rate limits are 60 requests/minute and 1,000/hour per token, with a 10,000-request monthly budget per account — every response carries rate-limit and budget headers, and your account page shows a per-endpoint usage breakdown.